cobekpa/nigerian-fintech-sentiment-annotation

Manual label of public sentiments of Fintech Apps (Opay, Moniepoint, Kuda, PalmPay) in Nigeria # 🇳🇬 Nigerian Fintech Sentiment Annotation ## 📌 Project Overview This project features the manual sentiment annotation of 572 user reviews from top-tier Nigerian fintech platforms: Opay, Moniepoint, Palmpay, and Kuda. The goal was to transform raw, unstructured Play Store feedback into a "Golden Dataset" optimized for sentiment analysis and NLP model training. This project demonstrates a full end-to-end data pipeline, from Docker-based tool deployment to final data validation. **Dataset Size:** 572 **Language:** English (including Nigerian Pidgin nuances) --- ## 🎯 Objective - Classify each user review into **one of three sentiment categories**: 1. **Positive** ✅ – User expresses satisfaction, praise, or favorable experience. 2. **Neutral** ⚪ – User expresses a neutral or factual statement without strong sentiment. 3. **Negative** ❌ – User expresses dissatisfaction, criticism, or negative experience. --- ## 🛠 Annotation Tool - **Label Studio** (via Docker) - Each record was manually reviewed and labeled based on the **actual review content** (ignoring the rating score). --- ## ✍ Annotation Guidelines - Read the full review before labeling. - Assign **one sentiment per review**: Positive, Neutral, or Negative. - Ignore numerical ratings; focus on **review content**. - Remove emojis and extra whitespace in text for consistency. - If a review is ambiguous, assign **Neutral**. - Ensure **no duplicates** and all records are labeled. For full details, see: `annotation/annotation_guidelines.md` --- ## 📊 Sentiment Distribution Based on the final 572 labeled reviews: | Sentiment | Percentage | |-----------|------------| | Positive | 80.24% | | Negative | 14.51% | | Neutral | 5.24% | **Wordclouds for Positive and Negative Sentiments**: --- ## 🗂 Project Structure ```nigerian-fintech-sentiment-annotation/ ├── annotation/ │ ├── annotation_guidelines.md │ └── label_studio_config.xml ├── data/ │ ├── annotated/ │ │ ├── final_labeled_fin …